First Run

Choose between single-instance and multi-instance modes, then send your first message.

Two Ways to Run

Single Instance: jaw serve

jaw serve
# Opens browser at http://localhost:3457

You get a chat interface immediately. Type a message and your AI agent responds.

Multi-Instance: jaw dashboard

jaw dashboard
# Opens browser at http://localhost:24576

The dashboard shows all your instances. Click Start to launch an instance, or Perm to make it persist across reboots.

Sending Your First Message

Type in the chat input and press Enter. The AI agent will respond using your configured CLI (Claude Code by default).

Switching CLI / Model

/cli claude    # Switch to Claude Code
/cli codex     # Switch to Codex
/cli gemini    # Switch to Gemini CLI
/model claude-opus-4-8  # Change model

Instance Controls

• Start — Launch instance (runs while dashboard is open)
• Stop — Stop instance
• Perm — Persist via launchd/systemd (survives reboots)

Next Steps

• Architecture — Understand the system design
• Employees — Dispatch tasks to other AI CLIs
• Dashboard Guide — Master the multi-instance UI